Publications by authors named "Yuichi Mori"
Article Synopsis
- The study focuses on the need for an objective method to evaluate and compare different computer-aided detection (CADe) algorithms used in colorectal cancer screening, as their performance varies and no standard exists.
- A modified Delphi approach was employed, where 25 experts generated and prioritized scoring criteria over eight months, ultimately identifying six key metrics, including sensitivity and adenoma detection rate.
- The resulting criteria aim to guide the development and improvement of CADe software, with future research suggested to validate these metrics on benchmark video datasets.
Article Synopsis
- AI has the potential to improve gastrointestinal endoscopy, but standardized methods are needed for its effective adoption in clinical practice.
- The QUAIDE Explanation and Checklist was created by a panel of 32 experts to provide guidelines for designing and reporting AI studies in this field.
- Consensus was achieved on 18 recommendations across key areas including data collection, outcome reporting, experimental setup, and result presentation, aiming to enhance research consistency and facilitate the use of AI in clinical settings.
Article Synopsis
- The study explores the use of computer-aided diagnosis (CADx) in the resect-and-discard strategy for the optical diagnosis of diminutive polyps during colonoscopy, aiming to improve diagnosis and reduce unnecessary pathology assessments.
- It involved a systematic review of existing research to analyze the effectiveness of CADx systems compared to traditional histology for small polyps (≤5 mm), including comparisons of CADx-assisted and unassisted methods.
- The meta-analysis included 11 studies with a total of 7400 polyps examined, highlighting the potential benefits and harms of using CADx in terms of accurate diagnosis and avoidance of false positives/negatives.
Article Synopsis
- - The study investigates how the location of colorectal polyps (proximal vs. distal colon) affects the performance of computer-aided diagnosis (CADx) in identifying neoplastic and non-neoplastic polyps during colonoscopies for polyps sized 5 mm or smaller.
- - It analyzes data from 11 studies involving 7,782 polyps, showing that CADx has lower specificity and accuracy in the proximal colon compared to the distal colon, while sensitivity is similar between both locations.
- - The conclusion highlights that while CADx works well for distal polyps, its inadequate performance in the proximal colon means it should not currently be used for those lesions until improved systems are created.
Article Synopsis
- A study evaluated the effectiveness of ChatGPT in answering questions about colorectal cancer (CRC) screening to improve patient awareness and adherence to screening programs.
- 15 CRC screening questions were assessed by experts and patients, focusing on accuracy, completeness, and comprehensibility.
- ChatGPT performed well, especially with patients, but highlighted the need for further training to align its responses with scientific evidence and guidelines.
Article Synopsis
- The study aimed to develop and validate an AI prediction system for assessing the risk of lymph node metastasis (LNM) in patients with T2 colorectal cancer (CRC), as traditional surgical approaches struggle with risk stratification.!* -
- Data from over 700 patients was analyzed, revealing that the AI model had a moderate prediction performance with a sensitivity of 97.8% but a low specificity of 15.6%, indicating many false positives in LNM predictions.!* -
- While the AI model shows promise for predicting LNM using basic clinical and pathologic data, improvements in accuracy could be achieved by training it with a larger and more diverse patient population from both Eastern and Western medical centers.!*
Article Synopsis
- - Significant advances in managing large laterally spreading tumors (LSTs) in colorectal polyps exist, but management practices vary globally, prompting the need for an international consensus.
- - A Delphi study involving 43 experts from 18 countries led to 42 statements reaching consensus regarding training, evaluation, resection techniques, and post-resection care for LSTs.
- - The resulting expert consensus aims to standardize practices and provide clear guidance for evaluating, resecting, and following up on LSTs worldwide.
Article Synopsis
- * The research involved 110 patients in remission and analyzed data from over 74,000 colonoscopy images, tracking patients for a year to see how well AI could predict relapse events compared to traditional methods.
- * Results showed that higher AI-based MES scores were linked to a significantly increased relapse rate, and the AI system improved consistency in diagnostic assessments made by non-specialist endoscopists.
Article Synopsis
- Computer-aided diagnosis (CADx) aims to improve the prediction of polyp histology during colonoscopy, potentially decreasing unnecessary removals of harmless polyps, though its overall benefits and risks remain uncertain.
- The study sought to evaluate the effectiveness of CADx for diagnosing small rectosigmoid polyps (≤5-mm) by comparing the accuracy of endoscopists' predictions with and without CADx assistance.
- Analysis of ten studies involving over 3,600 patients indicated that while CADx showed high sensitivity (87.3%) and specificity (88.9%) in identifying neoplastic polyps, there was no significant change in the rate of nonneoplastic polyps predicted to be avoided for removal when CADx
Article Synopsis
- - Integrating artificial intelligence (AI) into the study and treatment of inflammatory bowel disease (IBD) could significantly improve how doctors assess and predict disease activity through precise evaluations and standardised scoring methods
- - AI can support a comprehensive approach by combining data from endoscopy, histology, and other omics, which could lead to more personalised treatment options for IBD patients
- - Despite its potential, challenges such as data quality, ethical issues, and the need for standardised guidelines need to be addressed to successfully implement AI in clinical settings and research for IBD
Background And Aims: The American Society for Gastrointestinal Endoscopy (ASGE) AI Task Force along with experts in endoscopy, technology space, regulatory authorities, and other medical subspecialties initiated a consensus process that analyzed the current literature, highlighted potential areas, and outlined the necessary research in artificial intelligence (AI) to allow a clearer understanding of AI as it pertains to endoscopy currently.
Methods: A modified Delphi process was used to develop these consensus statements.
Results: Statement 1: Current advances in AI allow for the development of AI-based algorithms that can be applied to endoscopy to augment endoscopist performance in detection and characterization of endoscopic lesions.
Article Synopsis
- Accurate polyp size estimation is important for clinical decisions, and a new laser-based virtual scale endoscope (VSE) aims to improve measurement accuracy during colonoscopy compared to traditional visual assessment (VA) methods.
- In a study with 10 expert endoscopists reviewing video sequences of 90 polyps, VSE showed a relative accuracy of 75.1%, outperforming VA at 65.0% and snare-based estimation at 62.0%.
- The VSE significantly reduced errors in classifying smaller polyps as larger and vice versa, indicating it is more effective than VA or snare methods for accurate size estimation.
Article Synopsis
- * Studies have shown that CADx tools have high sensitivity and specificity in detecting neoplastic (cancerous) changes in these polyps.
- * The article reviews research data from prospective trials to assess if CADx tools meet the necessary standards for specific diagnosis strategies like "diagnose-and-leave," "resect-and-discard," and "DISCARD-lite."
Article Synopsis
- Trust and balanced reliance on digital resources are essential for accurate optical diagnoses and effective use of computer-aided diagnosis (CADx) in colonoscopy.
- Active learning with diverse polyp image datasets is beneficial for developing precise CADx systems, alongside enhancing doctors' digital literacy.
- Explainable AI improves interpretability of findings, helps communicate uncertainties to doctors, and fosters better collaboration by focusing on algorithm performance, transparency, and enhancing diagnostic skills.
Article Synopsis
- The systematic review analyzed the effects of computer-aided detection (CADe) in real-world colonoscopies, comparing it to standard procedures in nonrandomized studies.
- Results showed no significant improvement in adenoma detection rates or mean adenomas per colonoscopy when CADe was used, indicating its limited effectiveness in practical settings.
- The evidence for outcomes was rated as very low certainty, raising concerns about the reliability of findings from previously controlled randomized trials.
Article Synopsis
- AI-assisted detection during colonoscopy can improve adenoma detection rates (ADRs), potentially reducing the number of missed adenomas, but may lead to more nonneoplastic polyps being removed.
- A systematic review and meta-analysis of 21 randomized trials involving over 18,000 patients showed a significant increase in ADRs with AI assistance compared to standard colonoscopy.
- The study found that while the AI group detected more polyps, the extra time spent during the procedure was minimal, suggesting a balance between improved detection and the risk of overdiagnosis.
Article Synopsis
- A study was conducted to examine whether cancer screening tests actually improve longevity, focusing on life-years gained from such screenings compared to no screening.
- The analysis reviewed data from randomized clinical trials that included over 2 million participants and compared various common screening methods for cancers like breast, colorectal, lung, and prostate.
- Results showed varied follow-up durations across tests, but the key takeaway was that the real benefit in terms of added lifetime from these screenings remains unclear.
Computer-aided detection in real-world colonoscopy: enhancing detection or offering false hope?
Lancet Gastroenterol Hepatol
August 2023
Article Synopsis
- A "resect-and-discard" strategy for small polyps in the colon is challenging due to difficulty in distinguishing low-grade adenomas from more serious lesions like high-grade adenomas or cancers.
- This study evaluated the use of endocytoscopy (EC) with a specific indicator (normal pit-like structure, or NP-sign) to improve real-time diagnosis of low-grade adenomas compared to traditional imaging methods.
- Results showed that EC with NP-sign significantly outperformed other diagnostic techniques in terms of accuracy and specificity, making it a more reliable option for identifying which lesions can be safely discarded.